molecular

ACTIN-Molecular

ACTIN-Molecular interprets molecular results and maps to the datamodel described below. This molecular model is written to a per-sample json file. Currently, ACTIN-Molecular only supports interpretation of ORANGE results as produced by HMF Platinum

The ORANGE interpreter application requires Java 11+ and can be run as follows:

java -cp actin.jar com.hartwig.actin.molecular.orange.OrangeInterpreterApplication \
   -orange_json /path/to/orange.json \
   -external_trial_mapping_tsv /path/to/external_trial_mapping.tsv \
   -output_directory /path/to/where/molecular_json_file_is_written

The following assumptions are made about the inputs:

• The ORANGE json is the json output from ORANGE

The external trial mapping enables mapping of external trial acronyms to those used within ACTIN, in case there is duplication of trials across both sources.

Molecular Datamodel

Every sample, uniquely defined by their sample ID, has a molecular record with the following data:

Field	Example Value	Details
type	WGS	The type of molecular experiment done. Currently only 'WGS' is supported.
date	2022-01-14	The date on which the molecular results were obtained (optional field).
containsTumorCells	true	If false, implies that the tumor cell percentage in the biopsy was lower than the lowest detectable threshold.
hasSufficientQuality	true	If false, implies that the quality of the sample was not sufficient (e.g. too much DNA damage).

1 molecular characteristics

Field	Example Value	Details
purity	78%	The percentage of cells in the sequenced biopsy that originated from the tumor.
predictedTumorType	Melanoma (87%)	The tumor type of origin predicted based on the molecular data along with a likelihood.
isMicrosatelliteUnstable	false	If true, sample is considered microsatellite unstable. Can be left blank in case experiment does not determine MSI.
isHomologousRepairDeficient	false	If true, sample is considered homologous repair deficient. Can be left blank in case experiment does not determine HRD.
tumorMutationalBurden	14.2	Number of mutations in the genome per Mb. Can be left blank in case experiment does not determine TMB.
tumorMutationalLoad	115	Number of missense mutations across the genome. Can be left blank in case experiment does not determine TML.

N variants

Field	Example Value	Details
event	APC D1174fs	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	HIGH	Either HIGH, MEDIUM or LOW
gene	APC	The gene impacted by the variant
impact	p.D1174fs	The impact of the variant on the gene
variantCopyNumber	3.8	The number of copies of this variant in the tumor
totalCopyNumber	4.0	The total number of copies in the tumor on the mutated position
driverType	BIALLELIC	Either HOTSPOT, BIALLELIC or VUS
clonalLikelihood	100%	Likelihood that the variant exists in every tumor cell (is clonal).

N amplifications

Field	Example Value	Details
event	MYC amp	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	HIGH	Either HIGH, MEDIUM or LOW
gene	MYC	The gene that has been amplified
copies	150	Number of copies of this gene in the tumor
isPartial	false	Indicates whether the gene has been partially or fully amplified in the tumor

N losses

Field	Example Value	Details
event	TPS del	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	HIGH	Either HIGH, MEDIUM or LOW
gene	TP53	The gene that has been lost in the tumor
isPartial	true	Indicates whether the gene has been partially or fully lost in the tumor

N homozygous disruptions

Field	Example Value	Details
event	PTEN disruption	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	HIGH	Either HIGH, MEDIUM or LOW
gene	PTEN	The gene that has been homozygously disrupted in the tumor

N disruptions

Field	Example Value	Details
event	BRCA1 disruption	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	LOW	Either HIGH, MEDIUM or LOW
gene	BRCA1	The gene that has been disrupted
type	DUP	Type of disruption
junctionCopyNumber	1.1	Number of copies affected by this disruption
undisruptedCopyNumber	1.8	Remaining number of copies not impacted by this disruption
range	Intron 12 downstream	Provides additional details about the range impacted by the disruption.

N fusions

Field	Example Value	Details
event	EML4-ALK fusion	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	HIGH	Either HIGH, MEDIUM or LOW
fiveGene	EML4	The gene that makes up the 5' part of the fusion
threeGene	ALK	The gene that makes up the 3' part of the fusion
details	Exon 2 - Exon 5	Additional details about the fusion
driverType	KNOWN	Either KNOWN or PROMISCUOUS

N viruses

Field	Example Value	Details
event	HPV positive	A single representation of the event, expected to match with the event string from evidence section.
driverLikelihood	HIGH	Either HIGH, MEDIUM or LOW
name	Human papillomavirus type 16	Name of the virus found in the tumor sample
integrations	3	Number of integrations of detected virus in the tumor sample

N HLA alleles (with a single isReliable boolean indicating whether the HLA results are reliable)

Field	Example Value	Details
name	A*02:01	Name of the HLA allele
tumorCopyNumber	1.2	The number of copies of this HLA allele in the tumor sample.
hasSomaticMutations	false	A boolean indicating whether any mutations have occurred in this allele in the tumor

N pharmaco

Field	Example Value	Details
gene	DPYD	The gene for which the pharmaco entry is applicable
haplotype	1* HOM	Haplotypes found for the gene
haplotypeFunction	Function impact of corresponding haplotype

N wild-type genes (optional)

If the molecular experiment was not suitable to call wild-type genes this list can be omitted.

N actin trial evidences (with a single configured source name)

Field	Example Value	Details
event	BRAF V600E	The molecular event against which the evidence has been matched.
trialAcronym	Trial A	The acronym of the trial for which the evidence holds.
cohortCode	A	The ID of the cohort for which the evidence holds (optional, in case evidence holds for all cohorts within a trial).
isInclusionCriterion	1	Whether the evidence is involved in an inclusion or exclusion criterion for this trial.
type	ACTIVATED_GENE	The type of molecular event required by the trial. Can be SIGNATURE, ACTIVATED_GENE, INACTIVATED_GENE, AMPLIFIED_GENE, FUSED_GENE, MUTATED_GENE, WILD_TYPE_GENE and HLA_ALLELE
gene	BRAF	The gene required to be mutated by the trial (optional, empty in case of SIGNATURE and HLA_ALLELE)
mutation	null	The mutation required to be present by the trial (optional, only in case of MUTATED_GENE and HLA_ALLELE).

N external trial evidences (with a single configured source name)

Field	Example Value	Details
event	High TMB	The molecular event against which the evidence has been matched again.
trial	Trial A	The name of the trial for which the evidence holds.

N treatment evidences of various types, with a single configured source name. The types are as follows:

Type	Details
approvedEvidence	Treatments which are approved for the sample tumor type.
onLabelExperimentalEvidence	Treatments which are experimental for the specific sample tumor type.
offLabelExperimentalEvidence	Treatment which are experimental for a different tumor type.
preClinicalEvidence	Treatments that have some evidence but haven't reached experimental state yet.
knownResistanceEvidence	Evidence of resistance for treatments that are considered reliable.
suspectResistanceEvidence	Evidence of resistance for treatments that has not yet been confirmed to be reliable.

Interpretation of ORANGE results

The ORANGE interpreter application maps the ORANGE output to the molecular datamodel as follows:

Field	Mapping
sampleId	The ORANGE field sampleId
type	Hard-coded to WGS
date	The ORANGE field experimentDate
containsTumorCells	The PURPLE field hasReliablePurity
hasSufficientQuality	The PURPLE field hasReliableQuality

The characteristics are extracted as follows:

Field	Mapping
purity	The PURPLE field purity
predictedTumorOrigin	The CUPPA best cancer-type prediction along with the likelihood
isMicrosatelliteUnstable	The interpretation of PURPLE microsatelliteStabilityStatus
isHomologousRepairDeficient	The interpretation of CHORD hrStatus
tumorMutationalBurden	The PURPLE field tumorMutationalBurden
tumorMutationalLoad	The PURPLE field tumorMutationalLoad

The drivers are extracted from the following algorithms:

Driver Type	Algo	Details
variants	PURPLE	Union of reported somatic variants and reported germline variants.
amplifications	PURPLE	Union of reported somatic full gains and reported somatic partial gains.
losses	PURPLE	Union of reported somatic full losses and reported somatic partial losses.
homozygousDisruptions	LINX	All reported homozygous disruptions.
disruptions	LINX	All reported somatic gene disruptions.
fusions	LINX	All reported fusions with driver type mapped to either KNOWN or PROMISCUOUS.
viruses	VirusInterpreter	All reported viruses.

Note that all floating point numbers are rounded to 3 digits when ingesting data into ACTIN:

• variants: variantCopyNumber, totalCopyNumber, clonalLikelihood
• disruptions: junctionCopyNumber, undisruptedCopyNumber

The pharmaco entries are extracted from PEACH.

The wild-type genes are extracted from the ORANGE list of wild-type genes. In case the ORANGE results were not reliable, no wild-type genes will be set.

The evidence is extracted from PROTECT in the following steps:

1. Evidence is filtered for applicability based on ACTIN's internal applicability model. This applicability model removes evidence from any source that is never considered to be applicable.
2. Reported evidence is used exclusively for the inputs to the steps defined below

For ACTIN trial evidence, the following mapping is performed based on the assumption that PROTECT has been run against a SERVE database containing all ACTIN molecular criteria:

• trialAcronym and cohortId are extracted from the treatment field in PROTECT
• isUsedAsInclusion is set to true in case the evidence from PROTECT is responsive
• type, gene and mutation are extracted from the PROTECT ACTIN source data

For external trial evidence, treatments are mapped to trials from the ICLUSION source in PROTECT

For treatment evidence, the following categorization is done based on evidence from the CKB source in PROTECT:

Field	Filter
approvedEvidence	A-level on-label non-predicted responsive evidence.
onLabelExperimentalEvidence	Union of A-level that is either predicted or off-label responsive with B-level on-label non-predicted responsive evidence.
offLabelExperimentalEvidence	B-level off-label non-predicted responsive evidence.
preClinicalEvidence	All responsive evidence that is neither approved nor experimental.
knownResistanceEvidence	A or B-level non-predicted resistance evidence for a treatment for which non-preclinical evidence exists with equal or lower evidence level compared to the resistance evidence level.
suspectResistanceEvidence	Any other resistance evidence for a treatment with evidence with equal or lower evidence level compared to the resistance evidence level.

Note that in case of no tumor cells, both evidence and drivers are wiped empty. Any driver (and related evidence) is considered to be unreliable in such a case.

Version History and Download Links

• Upcoming (first release)